【作者】 谭小红；

【导师】 王雷刚；

【作者基本信息】 江苏大学 ， 材料学， 2013， 博士

【摘要】 多腔模细长杆类零件产品质量品质的不一致是笔类企业难以克服的技术瓶颈,设计一种基于平衡布置的一模四腔圆珠笔笔弹模具,借鉴传统注塑成型中熔体充模流动理论的研究方法,引入合理的假设与必要的简化,构建了细长杆注塑成型中熔体充模流动的数学模型。在此基础上分析了平衡设计的流道中,引起不平衡填充的各种因素,尤其是剪切力导致熔体在平衡流道中不平衡流动的机理。基于数值模拟技术以Y向最大变形作为流动不平衡的评价目标,通过改变模具温度、熔体温度、注射时间、保压压力等工艺参数,使得流动达到可以接受的平衡,并通过实验验证了理论分析和优化设计的结果,提出通过优化工艺参数组合直接或间接地优化多腔模产品质量的思路。本文确定相关产品质量的两个冲突目标—最大体积收缩率R1和最大轴向变形R2作为优化对象,首先比较了普通正交设计和基于噪音比的噪音设计在通过极差和方差设计在寻求最优工艺参数水平组合和因素显著性的差异,通过实验证明,运用稳健技术的Taguchi方法在单目标问题上不论从AVONA分析显著性,还有工艺参数最优水平组合,都优于普通的正交实验。相互冲突的双目标运用直观分析折中后的方案依然是稳健技术的Taguchi方法较优。在分析精度和各因素之间的交互作用上,提出基于CCD中心组合设计通过细长杆的响应面(RSM)回归模型,建立起设计变量与设计响应之间的非线性关系,求出了连续空间的最优工艺参数组合,并基于加权进行双目标的优化。与基于噪音比的正交实验相比较,响应面的精度更高,通过对细长杆Y向最大变形考察填充不平衡,基于响应面的最优工艺参数的填充平衡性更好。发展了一种基于遗传算法和神经网络的集成优化设计方法,使用GA优化BP神经网络的权值和阈值,种群中的每个个体都包含了一个网络所有权值和阈值,个体通过适应度函数计算个体适应度值,个体对应的最优适应度值由GA算法通过选择、交叉和变异操作找到BP神经网络预测用GA所得最优个体对网络初始权值和阈值赋值,网络经训练后预测函数输出。优化结果表明：遗传算法和神经网络集成的优化设计收敛速度快,对全局最优解的逼近程度高,可以迅速给出塑料制件期望值的工艺参数组合,通过训练好的网络获得了多腔模细长杆选用不同成型材料时模具温度、熔体温度、保压时间和保压压力等工艺参数对最大体积收缩率和最大轴向变形的影响规律。最后,针对注塑成形优化普通存在的多目标特性,首次提出了基于灰色关联分析(GCA)的多腔模细长杆注塑成形多目标优化方法。针对望目、望大、望小特征的目标函数,提出了不同的目标数据预处理方法。为了得到最佳的优化参数组合,对多目标的灰色关联度进行方差分析,并建立了显著影响变量的响应函数,从而得到最佳值。改进了传统的多目标灰色关联度寻优方法,研究结果表明,利用灰关联和理想解法的注射工艺多目标优化的改进优化措施进行多目标优化,可以获得最佳效果。模拟和优化的结果经过企业的实验验证,与生产实际吻合。更多还原

【Abstract】 The quality inconsistency of multi-cavity pen rod parts is the technical weakness for the pen companies to overcome. Based on the balanced arrangement, this paper designed a kind of one-mold-four-cavity ballpoint pen rod mold, which referred research methods melt filling flow theory in the traditional of the injection molding, introduced reasonable assumptions and necessary simplifications, and constructed the mathematical model of melt filling flow in the slender pen rod injection molding. Various factors of the unbalance filling in the runner in the balanced design were analyzed, especially the mechanism of shear-induced imbalance in the balanced flow channels. In the numerical simulation, Y-direction maximum deformation was determined as the evaluation objective of mobile unbalance. Through changing the process parameters such as mold temperature, melt temperature, injection time, holding pressure and so on, the flow filling and Y-direction maximum deformation reached the acceptable balance. On this basis, the theoretical analysis and design optimization results were experimentally proposed.It is determined two conflicting objectives on the quality of related product—the maximum volume shrinkage(R1) and the maximum axial deformation(R2) as the optimization objects, and the difference of normal orthogonal design and the Taguchi design based on the noise ratio was compared to seek the optimal process parameter combination and factor significance through the range and variance analysis. The experiment can show that Taguchi method with robust technology was better than the common orthogonal experiment at the single-objective problem no matter from AVONA analysis significance or the process parameter optimal level combination. After using the intuitive analysis, the conflicting bi-objective still preferred to use the Taguchi method with robust technology.Concerning the analysis on the interaction between the accuracy and factors, it was proposed to design the response surface (RSM) regression model based on Central composite design (CCD) of slender pen rod, the non-linear relationship was established between the design variable and design response, the optimal process parameter combination in the continuous space was sought, and the bi-objective optimization based on weighing was conducted. Compared with the orthogonal experiment based on the noise ratio, the accuracy of RSM was higher. With the inspection on the unbalance analysis of the Y-direction maximum deformation, the filling balance of the optimal process parameter based on RSM was better.A kind of integrated optimization design method based on the genetic algorithm and neural network was developed. GA was used to optimize the weight and threshold of BP neural network:Each individual in the population contains a BP network ownership value and threshold, and the individual calculates the fitness value through the fitness function. The individual corresponding optimal fitness value of BP network was found by GA algorithm selected, crossed and mutated. The network initial weight and threshold on the best individual assigned BP were obtained with GA. After training, the BP network would predict the function output. The optimization result can show that the optimized design method integrated by the genetic algorithm and neural network (GA-BP) has a fast convergence speed, higher approaching degree on the global optimal solution. It can quickly offer the process parameter combination of expected plastic part value. With the trained network, the influence rules of mold temperature, melt temperature, holding time and holding pressure and other process parameters on^l and^2were obtained when the multi-cavity mold slender rod was selected different materials.Finally, aiming at the multi-objective feature commonly existed in the injection molding and optimization, the multi-cavity slender pen rod injection molding multi-objective optimization method based on the analysis of gray correlation analysis (GCA) was firstly proposed. Concerning the objective functions of target-the-better, larger-the-better and smaller-the-better features, different target data preprocessing methods were proposed. In order to obtain the best optimal parameter combination, it conducted the variance analysis on the multi-objective gray correlation degree, established the response function significantly influencing the variable, and obtained the best value. The improved traditional research result can show that conducting the multi-objective optimization with the improved optimized measures on the injection process multi-objective optimization by use of the gray correlation analysis and ideal solution can obtain the best result.After simulation and verification of enterprises experimental, the results were consistent with the actual production.更多还原